Driving ROI in a multi-everything marketing reality


MROI means Marketing Return on Investment – or in other words: Which impact do the financial resources put into marketing have on sales?  This might sound like dry mathematical finance, but it is worthwhile for every CMO to engage in this topic for two reasons:

  • The term brings the idea of marketing as an investment in (future) turnover into focus – just as an investments into R&D – this alone is already a value in itself!
  • The budget of CMOs that do not consider their MROI will always be vulnerable: If marketing can’t proof its ROI, it remains a pure cost center!

Nevertheless, the challenge of a clean MROI measurement is in most cases the following: Everything happens at the same time. See figure 1 for illustration: What can the observed increase in sales in the highlighted month be attributed to? Your media spending? The raise of your promo intensity? Maybe it was the price reduction? Or maybe in the end, it was a factor outside of your company’s control, like e.g. a moment of weakness of your competitors or a rainy month that lead customers to the malls instead of the beach?

Figure 1


MMMs learn from the developments of the past. Since no past month equals another, modern statistics can identify patterns in the data and isolate the actual effects of particular marketing activities or media channels. This way, you can find out how strongly for example TV advertisement directly impacts sales in the short run and if this effect is reinforced by simultaneous online advertisement. In order to carry out a clean measurement of MROI one cannot evade statistic models. For the most part, these are referred to as Marketing Mix Models (MMMs) and in principle they do the following (see figure 2): they determine the correlation between “x” – various contemporaneously acting factors influencing sales – and the performance variable “y”, thus e.g. the development of sales over the course of the past years.


Figure 2


An in-house data science department is not necessarily needed to implement MMMs as there are sophisticated software solutions and service providers with a strong focus on consulting and supporting organizations in this process. In addition, the frequently expressed concern of models being a methodological “black box”, which someone needs to trust rather blindly, can be dispelled – quite the contrary: Good MMMs leave you with a transparent view of the underlying principles applied to determine the effectiveness of marketing activities.

But no MMM can work without data – both for “x” as well as “y”. Typically, at least data for the past 2-3 years on a weekly basis is required – the more data and the more granular, the better the models. MMMs solely work on aggregated data (per day, per week, per month), and there is no need for consumer-level data let alone personal information like names of individuals – good news in times of GDPR. Some of the data needed can even be obtained from public sources or panels, but e.g. your historical spending is best known to organizations themselves. The (very) pleasant side effect of a comprehensive MMM project therefore is the collection and cleaning of a database of marketing and sales data in one place, which otherwise would be scattered in silos across the organization or even saved in hard-to-handle formats like PDFs.

And the payback of the initial investment for an MMM project is demonstrated to be worthwhile: For who knows the correlation between “x” and “y” is able to adjust future marketing spending to move things in the right direction – independent scientific sources verify a potential increase in MROI of 10-30%.